Possible patch for tree-optimization/13000

[Richard Henderson]

On Wed, Jan 19, 2005 at 05:10:48PM -0500, Ian Lance Taylor wrote:
> I hope I'm not going down a rathole here, though.

Hard to say...  :-(  Unfortunately the Correct Way to do this is
queued for 4.1, where we build a CFG for functions before inlining.

> 	* gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
> 	branch if we don't need one.

Ok, except,

Please add a word of commentary about why we're bothering to do this
here, and why we can't use block_may_fallthru.  (At minimum, the 
behaviour of SWITCH_EXPR isn't the same until gimplification is 
complete.  Perhaps it's a mistake to give the two behaviours the
same name...)



r~